Decision trees, a non-parametric supervised learning method, are popular algorithms used for predictive modeling. The most well-known decision tree algorithms include the iterative dichotomizer (ID3), C4.5, CART, and C5.0. ID3 is only applicable for categorical features. C4.5 is an improvement on ID3 and has the ability to handle missing values and continuous attributes. The tree-growing process involves finding the best split at each node using the information gain. However, the C4.5 algorithm converts a continuous attribute into a dichotomous categorical attribute by splitting at a suitable threshold value that can produce maximum information gain.
Leo Breiman, a distinguished statistician, introduced a decision tree algorithm called the Classification and Regression Tree (CART). CART, unlike ID3 and C4.5, can produce decision trees that can be used for both...
